The work
You listen to and read AI-generated musical output — primarily Japanese lyrics, often paired with generated audio — and score it against a written rubric supplied by the AI lab. A typical task might ask you to judge whether a verse's word choice sounds like something a working J-pop or hip-hop lyricist would actually write, whether the phrasing scans naturally against a melody, or whether a generated chorus is uncomfortably close to a published song you recognize. Similarity checks are part of the job: you'll be asked to compare outputs against existing catalog material and flag overlap in melody, hook, or lyrical structure.
Much of the value you provide is written justification. Ratings alone are cheap; the labs want a sentence or two explaining why a line reads as translated-sounding, why a regional expression is misused, or why a rhyme scheme collapses in performance. Instructions arrive in English and can be long and specific, so comfort reading dense guideline documents in English while working critically in Japanese is essential.
What the screen looks for
- Verifiable music credentials — songwriting credits, released work, composition or performance training, or a body of music journalism
- Genuine native or near-native Japanese, assessed live through Mercor's Bilingual Competency Interview conducted in Japanese
- Concrete knowledge of the contemporary Japanese scene: artists, sub-genres, how slang and dialect actually appear in lyrics
- Evidence you can apply someone else's rubric consistently rather than substituting personal taste
Logistics
Fully remote and async. Start is immediate, with the engagement running up to six months. Hours are flexible with no cap; most contributors settle near 20 hours per week. You'll need headphones or monitors adequate for critical listening. Observed pay for this listing sits at $35–62/hr, set by experience and screening outcome — a range, not a guarantee.
